feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Adds MiniMax AI as a new first-class LLM provider alongside OpenAI, Azure, Groq, Ollama, and Mistral.

MiniMax offers OpenAI-compatible API endpoints, so this integration reuses LangChain's ChatOpenAI with a custom base URL (https://api.minimax.io/v1), following the same pattern as the existing Ollama provider.

Changes

• shell_ai/main.py: Add minimax to APIProvider enum, MiniMax config variables (MINIMAX_API_KEY, MINIMAX_MODEL, MINIMAX_API_BASE), API key validation, and provider initialization with temperature clamping (min 0.01)
• README.md: Add MiniMax environment variables, config example, and usage instructions
• tests/test_minimax_provider.py: 21 unit tests covering enum registration, config defaults, temperature clamping, provider validation, and API key detection
• tests/test_minimax_integration.py: 3 integration tests verifying live API calls with MiniMax-M2.5 and MiniMax-M2.5-highspeed models

Configuration

export SHAI_API_PROVIDER=minimax
export MINIMAX_API_KEY=your_api_key_here

Or via ~/.config/shell-ai/config.json:

{
   "SHAI_API_PROVIDER": "minimax",
   "MINIMAX_API_KEY": "your_api_key_here",
   "MINIMAX_MODEL": "MiniMax-M2.5"
}

Available models: MiniMax-M2.5 (default, 1M context), MiniMax-M2.5-highspeed (204K context).

Test plan

• 21 unit tests pass (pytest tests/test_minimax_provider.py)
• 3 integration tests pass with live MiniMax API (MINIMAX_API_KEY=... pytest tests/test_minimax_integration.py)
• Manual: SHAI_API_PROVIDER=minimax MINIMAX_API_KEY=... shai list files in current directory

---

No new dependencies required — MiniMax uses the existing langchain-openai package.